Can ultrasound attenuation measurement be used to characterise grain statistics in castings?

Abstract

Industrial inspection protocols are qualified using mock-ups manufactured according to the same procedure as the plant part. For coarse-grained castings, known for their low inspectability, relying on mock-ups becomes particularly challenging owing to the variability of grain properties among components. Consequently, there is a keen interest in the capability to verify whether the grain size of the component under test matches the qualification specification in-situ. This paper investigates the potential of an attenuation measurement for assessing the ultrasonic inspectability of coarse-grained components using qualified procedures in a practical setting. The experimental part of the study focuses on an industrial Inconel 600 mock-up with spatially varying attenuation, measured across the entire sample in an immersion tank. Three zones with distinctly different attenuations were examined using metallography, which allowed for calculating classical grain size histograms and two-point correlation functions. For one of the zones, we synthesised the microstructure with the same statistical properties numerically and simulated the propagation of ultrasound using a grain-scale finite element model. The results showed good agreement with the experiment, and lead to several suggestions for the reasons for the discrepancy, the varying grain size statistics being the most likely. A parametric study, which followed, depicted the effect of the mean and standard deviation-to-mean ratio of the log-normal grain size distribution on the attenuation of ultrasound and its frequency dependence. Most notably, we demonstrated the known non-uniqueness of the relationship between the log-normal grain size distribution parameters and the attenuation. We suggested that the correlation length calculated from a single exponential fit to the two-point correlation function is a more robust metric describing grain statistics for this context, which can be obtained from attenuation. The correlation lengths estimated from measured attenuation using the second-order approximation model for the three zones of the studied mock-up yielded results of acceptable accuracy. We concluded that this metric could replace the average grain size in practical settings, as it retains more statistical information than the mean grain size and allows for linking measurements to the established theoretical attenuation models which this paper demonstrates.